Prediction of international rice production using long short-term memory and machine learning models

Abstract

Rice, a staple food source globally, is in high demand and production across the world. Its consumption varies in different countries, with each nation having its unique way of incorporating rice into its diet. Recognizing the global nature of rice, its production is a crucial aspect of ensuring its availability, agriculture forecasting, economic stability, and food security. By predicting its production, we can develop a global plan for its production and stock, thereby preventing issues like famine. This paper proposes machine learning (ML) and deep learning (DL) models like linear regression, ridge regression, random forest (RF), adaptive boosting (AdaBoost), categorical boosting (CatBoost), extreme gradient boosting (XGBoost), gradient boosting, decision tree, and long short-term memory (LSTM) to predict international rice production. A total of nine ML and one DL models are trained and tested on the international dataset, which contains the rice production details of 192 countries over the last 62 years. Notably, linear regression and the LSTM algorithm predict rice production with the highest percentage of R-squared (R2 ), 98.40% and 98.19%, respectively. These predictions and the developed models can play a vital role in resolving crop-related international problems, uniting the global agricultural community in a common cause.